P-664 Background: When it closed in 2003 the Britannia Zinc Smelter in Avonmouth was the largest source of atmospheric cadmium emissions in the UK and the biggest smelter of its kind in the world. Zinc production had been taking place at the site for over 70 years and during that time large amounts of cadmium were emitted to air. With close to 50,000 people living within 5km it is possible that these emissions may have led to significant cadmium exposure. Cadmium is a known nephrotoxin. The initial sign of cadmium induced renal damage is tubular proteinuria. There is increasing evidence that tubular damage may occur at low cumulative doses (U-Cd 1–3μg/g creatinine) such as those that might be expected from environmental exposure. The Avonmouth Pilot Study will investigate urinary cadmium concentrations and the prevalence of tubular proteinuria in the potentially exposed population. Since cadmium pollution in the vicinity of smelting operations occurs largely as result of emissions to air, the exposure classification for this study has been carried out using air dispersion modelling; the results of this modelling will be presented. Methods: ADMS-Urban 2.0 has been used to model emissions from up to 13 stacks from 1996–2002. Analysis has been carried out to explore the model's sensitivity to various parameters and to determine whether stack emissions alone can account for the concentrations recorded at the monitoring sites. The model has been validated using air monitoring data and has been used to define exposure groups. Results: Modelled concentrations show a significant correlation with those recorded at the monitoring sites (correlation significant at the 0.01 level). The model under predicts cadmium concentrations. Concentrations at the monitoring sites can not be accounted for by stack emissions alone; it is thought that fugitive emissions at the site may have been significant. The model shows sensitivity to various parameters including buildings and roughness length. Conclusion: Although under-predicting actual air concentrations, the modelled cadmium concentrations show a good correlation with monitor data. The modelled levels therefore provide useful estimates of potential cadmium exposure.Figure